构建首个大规模音频推理模型，提升语音理解逻辑能力
Audio-Reasoner: Improving Reasoning Capability in Large Audio Language Models
- 设计结构化思维链训练流程，生成高质量音频推理数据
- 在多个基准上达到领先表现，最高提升25.42%
- 适合需要深度音频分析的智能系统研发者
当前多模态推理研究严重忽视音频模态。我们提出Audio-Reasoner，一个面向音频任务的大型语言模型，具备深度推理能力。通过精心构建包含多样任务的大规模音频数据集，并采用闭源模型进行二次标注、问答生成及结构化思维链（COT）处理，形成包含120万条富含推理样本的高质量数据集CoTA。基于推理扩展原则，在CoTA上训练Audio-Reasoner，显著提升其音频推理逻辑能力。实验表明，在多个关键基准测试中均达领先水平：MMAU-mini提升25.42%，AIR-Bench chat/foundation分别提升14.57%和10.13%，MELD提升8.01%。研究强调结构化思维链训练对推动音频推理的核心作用。
原文摘要 · Abstract (English)
Recent advancements in multimodal reasoning have largely overlooked the audio modality. We introduce Audio-Reasoner, a large-scale audio language model for deep reasoning in audio tasks. We meticulously curated a large-scale and diverse multi-task audio dataset with simple annotations. Then, we leverage closed-source models to conduct secondary labeling, QA generation, along with structured COT process. These datasets together form a high-quality reasoning dataset with 1.2 million reasoning-rich samples, which we name CoTA. Following inference scaling principles, we train Audio-Reasoner on CoTA, enabling it to achieve great logical capabilities in audio reasoning. Experiments show state-of-the-art performance across key benchmarks, including MMAU-mini (+25.42%), AIR-Bench chat/foundation(+14.57%/+10.13%), and MELD (+8.01%). Our findings stress the core of structured CoT training in advancing audio reasoning.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。